Exploration of the Specific Pathology of HXMM Tablet Against Retinal Injury Based on Drug Attack Model to Network Robustness

Abstract

Retinal degenerative diseases are related to retinal injury because of the activation of the complement cascade, oxidative stress-induced cell death mechanisms, dysfunctional mitochondria, chronic neuroinflammation, and production of the vascular endothelial growth factor. Anti-VEGF therapy demonstrates remarkable clinical effects and benefits in retinal degenerative disease patients. Hence, new drug development is necessary to treat patients with severe visual loss. He xue ming mu (HXMM) tablet is a CFDA-approved traditional Chinese medicine (TCM) for retinal degenerative diseases, which can alleviate the symptoms of age-related macular degeneration (AMD) and diabetic retinopathy (DR) alone or in combination with anti-VEGF agents. To elucidate the mechanisms of HXMM, a quantitative evaluation algorithm for the prediction of the effect of multi-target drugs on the disturbance of the disease network has been used for exploring the specific pathology of HXMM and TCM precision positioning. Compared with anti-VEGF agents, the drug disturbance of HXMM on the functional subnetwork shows that HXMM reduces the network robustness on the oxidative stress subnetwork and inflammatory subnetwork to exhibit the anti-oxidation and anti-inflammation activity. HXMM provides better protection to ARPE-19 cells against retinal injury after H₂O₂ treatment. HXMM can elevate GSH and reduce LDH levels to exhibit antioxidant activity and suppress the expression of IL-6 and TNF-α for anti-inflammatory activity, which is different from the anti-VEGF agent with strong anti-VEGF activity. The experimental result confirmed the accuracy of the computational prediction. The combination of bioinformatics prediction based on the drug attack on network robustness and experimental validation provides a new strategy for precision application of TCM.

Keywords: age-related macular degeneration; diabetic retinopathy; drug attack; he xue ming mu tablet; inflammation; network robustness; oxidative stress.

FIGURE 1

Schematic diagram of quantitative algorithm of the drug attack on network robustness (A) and permutation test of HXMM attack to AMD and DR network (B). (A) Simulation of the disease network and the impact of drug attack on the robustness of the network. The schematic diagram simulates the calculation process of the impact of drug attack on the robustness of the disease network. In the same disease network, deleting different nodes will cause different changes in the network structure and robustness. Change of network topological features after drug attack was used as robustness index (RI) of networks. The null distribution of RI of 100 random networks was generated to maintain the same edge number of the true network and randomize interactions between nodes. The dotted area indicated the 95% confidence interval via the Permutation test. Compared with the null distribution, the statistical significance of RI of true network can be calculated and normalized RI can be corrected. We can normalize the p-value (observed statistics of RI) to z-score, which is (n-Normalized RI). By constructing a random network, simulating the overall distribution of the impact of deleted nodes on the robustness of the random network (permutation test), investigating the position of the drug on the real network disturbance in the overall distribution, calculating the corrected disturbance rate, and the intensity of drug disturbance on the real network can be evaluated objectively. (B) Permutation test of HXMM RIs of real network were shown. To evaluate the drug disturbance on the network from four topological features, including the AD, ASPL, CC, and BC. The red arrow pointed to the RI under the real network. The null distribution of RI of drug attack on the random network was shown as normal distribution curve. In AMD network, the normalized RI of AD, ASPL, DC and CC were -3.53, 4.64,-0.83 and 0.5, the corresponding figures of DR were -6.77, 7.01,-0.16.-11.88.

FIGURE 2

Major pathological process of AMD (A) and DR (B) and FDA-approved-drug disturbance on these pathological process subnetworks (C). (A,B) Network of enriched biological process GO terms of AMD and DR (left), GO terms clustered in one group were marked with the same color (red—oxidative stress, blue—inflammation, yellow—angiogenesis, green—blood coagulation, organ—extracellular matrix, and purple—neuro death). Node size changes with the count of genes in this GO term. The circos plot of genes in GO terms cluster group shows the overlapping of major pathological processes. (C) FDA-approved-drug disturbance on these pathological process subnetworks. Anti-VEGF agents, growth factors, corticosteroids, angiotensin II receptor blockers, complements, immunosuppressants, anticoagulants, antidiabetic drugs were marked by blocks with different colors from red to gray.

FIGURE 3

Evaluation of the Robustness of disease network disturbance after multi-target drug attack. (A) Distribution of total scores for different pathological processes in AMD and DR. The scores in brackets indicated the ratio of HXMM total scores greater than those FDA-approved drugs under the same pathological process. (B) Multi-target drug attack prediction on topological scores for AMD. (C) Multi-target drug attack prediction on topological scores for DR. The color depth changed with the increase of the score. The red was a positive number, which represented a stronger positive disturbing effect, and the blue was a negative number, which meant that it had a relatively weak disturbing ability. Among them, HXMM had shown better scores in multiple processes, especially in oxidation stress, so we use the red square to circle it.

FIGURE 4

HXMM protects against H₂O₂-induced retinal injury in ARPE-19 cells (A) cell viability of viability of ARPE-19 cells treated with H₂O₂ and HXMM (n = 6), Changes in ARPE-19 cell morphology in response to HXMM and H₂O₂. APRE-19 cells without HXMM or H₂O₂ (control); APRE-19 cells with H₂O₂ (model); APRE-19 cells with 4 mg/ml and 2 mg/ml HXMM and then treated by 65 μM H₂O₂ (HXMM-H, HXMM-L) (B) fold change of process of oxidative stress, LDH activity (n = 4) and GSH level (n = 4) (C) fold change of process of inflammation, IL-6 activity (n = 4) and TNF-α level (n = 4) (D) fold change of process of angiogenesis, VEGFA level (n = 4) and VEGFB (n = 4). Data were expressed as the mean ± SD, #p < 0.05 versus control, *p < 0.05 versus model was considered significant, **p < 0.01 was considered significant statistical difference, ***p < 0.001 was considered extremely significant statistical difference.

FIGURE 5

Analysis of the content ratio and action intensity in different pathological processes of HXMM. (A) Hierarchy network of chemical compound, drug-attacked nodes in the AMD network, and DR-related pathological processes. This network was composed of the interaction relationship between HXMM’s intervention in AMD and DR disease-related potential compounds and the targets of oxidative stress, inflammation, and angiogenesis. The pink triangle represents the potential compound of HXMM, the red circle represented the special target of oxidative stress, and the blue circle represented the special target of inflammation. A circle composed of multiple colors indicates that the target exists in multiple processes at the same time. There were 9 targets (composed of red, blue, and yellow) shared by oxidative stress, inflammation, and angiogenesis, 19 targets (composed of red and blue) can be seen in oxidative stress and inflammation, and 3 targets (composed of red and yellow) shared by oxidative stress and angiogenesis. (B) The bubble plot of the sum of peak area ratios of target-related compounds involved in the angiogenesis, inflammation, and oxidative stress. The abscissa denotes the three pathological processes mentioned in panel 5A, and the ordinate is the target in panel 5A. The size of the bubbles is related to the ratio of the total peak area of the target-related compounds.